Designing a multi-legged robot as a test-bed for motion intelligence mechanisms

2008-10 till 2011-09
Research Areas: 

This project explores the foundations of legged locomotion from a high-level planning perspective and as a result will shed light on the link between reactive and cognitive systems. To approach this task, a novel hexapod walking machine – Hector – has been developed. This robot is the result of a multi-disciplinary approach that brings together engineers, biologists, materials scientists, designers and computer scientists. It features lightweight construction, bio-inspired elastic joint actuation, lightweight 2D body segment actuation, decentral control and finally planning capabilities.


Methods and Research Questions: 

A major field of research in cognitive robotics deals with the question, how robots may learn to reach distant points in an environment of such complexity that solely reactive behaviors are not sufficient. This question is especially interesting for legged robots since they have many more degrees of freedom than wheeled systems.

A major field of research in cognitive robotics deals with the question, how robots may learn to reach distant points in an environment of such complexity that solely reactive behaviors are not sufficient to reach the goal. This task is especially interesting for legged robots since they have more degrees of freedom than wheeled systems. For this project we have decided for a six-legged machine rather than a two-legged one. One reason is that in two-legged walking a maximum effort has to be invested into the dynamic stabilization of the robot already during simple walking on flat terrain. Careful operation (falling has to be prevented at any costs) impedes the exploration of an otherwise large parameter space. The hexapod robot has 22 DoF which allows him to overcome obstacles like stairs similarly to the biological example. To fully utilize the opportunities of legged locomotion, robust and fault-tolerant walking mechanisms in combination with planning abilities are needed. Until now, the neuronal mechanisms which underly walking have not been fully understood. Up to date walking machines often make their legs only follow pre-calculated trajectories and handle only predefined exceptions. Biologists know that this, at the best, is only mimicking of walking.

For the MULERO project a new hexapod robot systems had to be developed. This robot had to fulfill several requirements. First, it had to copy the most important body features of a stick insect which are the relative distances of the leg onsets, the orientation of joint axes, additional degrees of freedom in terms of body-joints, muscle-like elastic drive systems and a self-carrying exoskeleton. Second, the control of the robot had to comply with the decentral architecture which is found in the neuronal control centers of insects. Third, the control system had to maintain a modular structure to add the planning and reasoning systems which utilize internal body models. Fourth, the robot body had to be designed such that further bio-inspired proprioception can be easily integrated. Fifth, the robot has to serve as a versatile carrier for additional environmental sensors like antennae and vision systems. To realize this robotic system, new self-contained, lightweight joint drives (BioFlex Drives) have been developed which are able to generate muscle like elastic features. The lightweight construction of the housing was influenced by experts on functional design (Folkwang University Essen) and by polymer specialists (Leibniz Institute for Polymer Research, Dresden). Besides engineering research on high performance embedded drive and sensor systems biological research currently focuses on biomimetic transfer of sensor concepts from insects. Walking and planning principles are currently tested on a dynamics simulation of the new robot system.


Expected outcomes of the MULERO project are manifold. The physical outcome is a new type of hexapod robot with increased maneuverability, elastic actuation and torque reserves that enable it to negotiate cluttered environments. The basis for this robot is created by a new type of joint actuation system called BioFlex Drive. These drives possess integrated electronics and copy elastic features of muscle-driven systems. The self-contained character of this drive allows an easy setup of different limbs. The drives also contain a new sensorized elastomer coupling (patent pending). In terms of control-theoretical outcomes, the MULERO project has guided the development of new control approaches for muscle driven actuation. The coordination of walking based on elastic joints has been investigated in a dynamics simulation of the robot. Results show that implicit rather than explicit control rules as found in insects facilitate coordinated walking in different environments. Another expected outcome is the planning system which uses the low-level, implicit walking control and generates escape strategies from movement problems that are related to the local environment the robot is embedded in. Further biological research on the integration of rich sensor information into the walking control system has been triggered by the MULERO project. The developed robot is also the basis of the new EU EMICAB project.